Flexibility in the ABC transporter MsbA: Alternating access with a twist.
ABSTRACT ATP-binding cassette (ABC) transporters are integral membrane proteins that translocate a wide variety of substrates across cellular membranes and are conserved from bacteria to humans. Here we compare four x-ray structures of the bacterial ABC lipid flippase, MsbA, trapped in different conformations, two nucleotide-bound structures and two in the absence of nucleotide. Comparison of the nucleotide-free conformations of MsbA reveals a flexible hinge formed by extracellular loops 2 and 3. This hinge allows the nucleotide-binding domains to disassociate while the ATP-binding half sites remain facing each other. The binding of the nucleotide causes a packing rearrangement of the transmembrane helices and changes the accessibility of the transporter from cytoplasmic (inward) facing to extracellular (outward) facing. The inward and outward openings are mediated by two different sets of transmembrane helix interactions. Altogether, the conformational changes between these structures suggest that large ranges of motion may be required for substrate transport.
- SourceAvailable from: Robert Tampé[Show abstract] [Hide abstract]
ABSTRACT: The ATP-binding cassette (ABC) transporter associated with antigen processing (TAP) participates in immune surveillance by moving proteasomal products into the endoplasmic reticulum (ER) lumen for major histocompatibility complex class I loading and cell surface presentation to cytotoxic T cells. Here we delineate the mechanistic basis for antigen translocation. Notably, TAP works as a molecular diode, translocating peptide substrates against the gradient in a strict unidirectional way. We reveal the importance of the D-loop at the dimer interface of the two nucleotide-binding domains (NBDs) in coupling substrate translocation with ATP hydrolysis and defining transport vectoriality. Substitution of the conserved aspartate, which coordinates the ATP-binding site, decreases NBD dimerization affinity and turns the unidirectional primary active pump into a passive bidirectional nucleotide-gated facilitator. Thus, ATP hydrolysis is not required for translocation per se, but is essential for both active and unidirectional transport. Our data provide detailed mechanistic insight into how heterodimeric ABC exporters operate.Nature Communications 11/2014; 5:5419. · 10.74 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Despite the growing interest in membrane proteins, their crystallization remains a major challenge. In the course of a crystallographic study on the multidrug ATP-binding cassette transporter BmrA, mass spectral analyses on samples purified with six selected detergents revealed unexpected protein contamination visible for the most part on overloaded SDS-PAGE. A major contamination from the outer membrane protein OmpF was detected in purifications with Foscholine 12 (FC12) but not with Lauryldimethylamine-N-oxide (LDAO) or any of the maltose-based detergents. Consequently, in the FC12 purified BmrA, OmpF easily crystallized over BmrA in a new space group, and whose structure is reported here. We therefore devised an optimized protocol to eliminate OmpF during the FC12 purification of BmrA. On the other hand, an additional band visible at ∼110 kDa was detected in all samples purified with the maltose-based detergents. It contained AcrB that crystallized over BmrA despite its trace amounts. Highly pure BmrA preparations could be obtained using either a ΔacrAB E. coli strain and n-dodecyl-β-D-maltopyranoside, or a classical E. coli strain and lauryl maltose neopentyl glycol for the overexpression and purification, respectively. Overall our results urge to incorporate a proteomics-based purity analysis into quality control checks prior to commencing crystallization assays of membrane proteins that are notoriously arduous to crystallize. Moreover, the strategies developed here to selectively eliminate obstinate contaminants should be applicable to the purification of other membrane proteins overexpressed in E. coli.PLoS ONE 12/2014; 9(12):e114864. · 3.53 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: L-Arabinose sugar residues are relatively abundant in plants and are found mainly in arabinan polysaccharides and in other arabinose-containing polysaccharides such as arabinoxylans and pectic arabinogalactans. The majority of the arabinose units in plants are present in the furanose form and only a small fraction of them are present in the pyranose form. The L-arabinan-utilization system in Geobacillus stearothermophilus T6, a Gram-positive thermophilic soil bacterium, has recently been characterized, and one of the key enzymes was found to be an intracellular β-L-arabinopyranosidase (Abp). Abp, a GH27 enzyme, was shown to remove β-L-arabinopyranose residues from synthetic substrates and from the native substrates sugar beet arabinan and larch arabinogalactan. The Abp monomer is made up of 448 amino acids, and based on sequence homology it was suggested that Asp197 is the catalytic nucleophile and Asp255 is the catalytic acid/base. In the current study, the detailed three-dimensional structure of wild-type Abp (at 2.28 Å resolution) and its catalytic mutant Abp-D197A with (at 2.20 Å resolution) and without (at 2.30 Å resolution) a bound L-arabinose product are reported as determined by X-ray crystallography. These structures demonstrate that the three-dimensional structure of the Abp monomer correlates with the general fold observed for GH27 proteins, consisting of two main domains: an N-terminal TIM-barrel domain and a C-terminal all-β domain. The two catalytic residues are located in the TIM-barrel domain, such that their carboxylic functional groups are about 5.9 Å from each other, consistent with a retaining mechanism. An isoleucine residue (Ile67) located at a key position in the active site is shown to play a critical role in the substrate specificity of Abp, providing a structural basis for the high preference of the enzyme towards arabinopyranoside over galactopyranoside substrates. The crystal structure demonstrates that Abp is a tetramer made up of two `open-pincers' dimers, which clamp around each other to form a central cavity. The four active sites of the Abp tetramer are situated on the inner surface of this cavity, all opening into the central space of the cavity. The biological relevance of this tetrameric structure is supported by independent results obtained from size-exclusion chromatography (SEC), dynamic light-scattering (DLS) and small-angle X-ray scattering (SAXS) experiments. These data and their comparison to the structural data of related GH27 enzymes are used for a more general discussion concerning structure-selectivity aspects in this glycoside hydrolase (GH) family.Acta Crystallographica Section D Biological Crystallography 11/2014; 70(Pt 11):2994-3012. · 7.23 Impact Factor
Flexibility in the ABC transporter MsbA: Alternating
access with a twist
Andrew Ward*, Christopher L. Reyes†, Jodie Yu†, Christopher B. Roth†, and Geoffrey Chang†‡
Departments of *Cell Biology and†Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, CB-105, La Jolla, CA 92037
Communicated by K. Barry Sharpless, The Scripps Research Institute, La Jolla, CA, October 3, 2007 (received for review August 6, 2007)
ATP-binding cassette (ABC) transporters are integral membrane
proteins that translocate a wide variety of substrates across cel-
lular membranes and are conserved from bacteria to humans. Here
we compare four x-ray structures of the bacterial ABC lipid flip-
pase, MsbA, trapped in different conformations, two nucleotide-
bound structures and two in the absence of nucleotide. Compar-
ison of the nucleotide-free conformations of MsbA reveals a
flexible hinge formed by extracellular loops 2 and 3. This hinge
allows the nucleotide-binding domains to disassociate while the
ATP-binding half sites remain facing each other. The binding of the
nucleotide causes a packing rearrangement of the transmembrane
helices and changes the accessibility of the transporter from
cytoplasmic (inward) facing to extracellular (outward) facing. The
inward and outward openings are mediated by two different sets
of transmembrane helix interactions. Altogether, the conforma-
tional changes between these structures suggest that large ranges
of motion may be required for substrate transport.
lipid flippase ? membrane protein structure ? multidrug transport
couple ATP hydrolysis to biological work (1–6). ABC transporters
are integral membrane proteins that mediate the transport of
diverse substrates such as ions, lipids, peptides, metabolites, che-
motherapeutic drugs, and antibiotics across cellular membranes.
ABC transporters also play a role in multidrug resistance (7), and
mutations in these proteins have been shown to cause diseases such
as cystic fibrosis (8–10).
ABC transporters are minimally composed of two nucleotide-
binding domains (NBDs), which contain the ABC signature motifs
and two transmembrane domains (TMDs). The NBDs are the site
of ATP hydrolysis and are highly conserved among the ABC
transporter superfamily. In contrast, the TMDs, which form the
transport pathway mediating accessibility to either side of the
membrane, have more sequence diversity and vary in the total
number of transmembrane helices. Despite differences in both
genetic organization of the cognate domains (11) and substrate
specificity, known structures of ABC transporters and isolated
NBDs exhibit an internal twofold or pseudotwofold symmetry that
is formed between interacting monomers (12–24).
MsbA is found in Gram-negative bacteria and transports lipid A
facing) to the periplasmic leaflet (outward-facing) of the inner
membrane (25–31). Functional studies have also shown that MsbA
a wide spectrum of drug molecules (32–34). Additionally, MsbA
shares significant sequence identity with human multidrug resis-
tance protein 1 (MDR1) and LmrA from Lactococcus lactis, which
have been implicated in multidrug resistance (10, 35).
The crystal structure of the bacterial transporter Sav1866 in
complex with ADP revealed the high-resolution structure and
topology for this class of lipid and multidrug-resistance ABC
exporters (22). Previously published x-ray structures of MsbA were
inconsistent with the Sav1866 structure (36). Upon reexamination
of the data, we discovered an error in the assignment of the hand,
he ATP-binding cassette (ABC) is a highly conserved domain
found in a variety of prokaryotic and eukaryotic proteins that
software program that converted intensities to structure factors
from HKL2000 to PHASES format. This problem was not readily
apparent in the original analysis of the data from our crystals and
resulted in the building of incorrect models of MsbA (37–39). The
use of multicopy refinement procedures allowed us to obtain
using a single-copy refinement procedure (CNS1.2), we corrected
the two previously published apo-MsbA crystal structures (37, 38)
and solved two new nucleotide-bound structures. Comparison of
transporters (16, 22, 24) and reveals that MsbA can undergo large
conformational changes, which may be essential for transport.
Overview. The structures of MsbA in complex with 5?-adenylyl-?-
apo structures have two different inward (intracellular)-facing
conformations. To reflect the resolution of the data, only the 3.7-Å
a full model, whereas the others include only the C? positions. Fig.
the same topology as Sav1866, and the fold of the NBD agrees with
were determined outside of their native membrane environment,
the different conformations reveal motions within the MsbA dimer
that suggest a plausible mechanism by which accessibility switches
from inward- to outward-facing [supporting information (SI)
bacterial orthologs: Escherichia coli (MsbA-EC), Vibrio cholerae
(MsbA-VC), and Salmonella typhimurium (MsbA-ST) (SI Fig. 6).
their ATPase activity was assayed (see Materials and Methods) to
address previously raised concerns about the functional integrity of
protein used for crystallization (41). Purified detergent-solubilized
MsbA from all three species, MsbA-ST, -EC, and -VC, exhibited
significant ATPase activities with Vmaxvalues of 2.0 ?mol/min per
milligram, 1.7 ?mol/min per milligram, and 5.3 ?mol/min per
indeed higher than previously reported (29), confirming that all
proteins were active as isolated. The Kmvalues were 0.300, 0.440,
and 0.280 mM, respectively. The Vmax and Km values are quite
Author contributions: A.W. and C.L.R. contributed equally to this work; A.W., C.L.R., and
G.C. designed research; A.W., C.L.R., J.Y., C.B.R., and G.C. performed research; G.C. con-
tributed new reagents/analytic tools; A.W., C.L.R., J.Y., and G.C. analyzed data; and A.W.,
C.L.R., J.Y., C.B.R., and G.C. wrote the paper.
The authors declare no conflict of interest.
Data deposition: The coordinates and structure factors have been deposited in the Protein
Data Bank, www.pdb.org (PDB ID codes 3B5W, 3B5X, 3B5Y, 3B5Z, and 3B60).
‡To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
November 27, 2007 ?
vol. 104 ?
no. 48 ?
similar to values previously reported for MsbA [0.95 ?mol/min per
milligram and 0.3 mM, respectively (42)]. Furthermore, the ATP
analogue, AMPPNP, inhibited MgATP hydrolysis of MsbA-ST in
a competitive manner with a Kirange of 10–20 ?M (SI Fig. 7C).
MsbA-ST was also inhibited by vanadate in a concentration-
dependent manner with full inhibition observed at ?100 ?M (SI
Fig. 7D). All these binding parameters are in a range similar to the
homologous ABC transporters, MDR1 (43), LmrA (35), and
Structure Determination. Our structural analysis of MsbA began by
using data of a previously undescribed crystal form of MsbA from
S. typhimurium in complex with AMPPNP (MsbA-AMPPNP).
of the experimental protein phases and only three minor gaps per
monomer in otherwise continuous electron density (Fig. 2A). The
electron-density maps revealed a conformation similar to the
homologous (36% identical and 56% similar) bacterial ABC trans-
as a starting point for the interpretation of our experimental
electron-density map to preserve the helical registration and as-
signment of the C? positions. In all models, we were able to fit all
but the first 10 amino acids into our electron-density maps by
using mercury bound to cysteine as markers for topology where
The protein phases of the MsbA-AMPPNP crystal form were
determined by a multiple anomolous dispersion (MAD) experi-
ment by using PHASES (45) and the proper sign between the
anomalous Friedel pairs. Solvent-flattening and fourfold noncrys-
tallographic symmetry (NCS) averaging produced experimental
electron-density maps that were then used to manually position a
homology model of an MsbA-AMPPNP monomer (Fig. 2A) based
on the Sav1866 structure. The homology model was generated by
(46). A simulated annealing procedure (CNS1.2) using fourfold
NCS restraints and a final round of B-factor refinement with bulk
solvent correction (47) resulted in a single-copy model (Fig. 1A)
with an Rcrystof 30% and Rfreeof 34% at 4.5-Å resolution (SI Table
1). Hydrogen bond restraints were used for all secondary-structure
elements. The model was validated by generation of a composite
omit map (CNS1.2) (SI Movie 2), and the topology of the structure
was experimentally confirmed by the proximity of two mercury
atoms, which were used in the map-phasing calculations, to two
cysteine residues in each monomer (Fig. 2A). A full model was
refined, but the resolution did not allow for precise determination
of the side-chain positions.
Higher-resolution data at 3.7 Å were also collected on a related
crystal form of MsbA with AMPPNP (SI Table 1) and solved by
structure refined to a Rcrystof 32% and Rfreeof 34% (CNS1.2) and
is nearly identical with the model solved to 4.5 Å with an rmsd of
0.37 Å between the C? positions. The helix registration and the
position of the nucleotide were verified by using an averaged omit
map phased from a model truncated to polyalanine/polyglycine.
After fourfold NCS averaging, the omit map revealed the position
of several bulky side chains and confirmed the registration of the
helices (Fig. 2 B and C). In addition, the position of the nucleotide
and the neighboring Tyr-351 is resolved in the map (Fig. 2D). This
model was also validated by the composite omit map (SI Movie 3).
This model was used as the starting point for interpretation of the
electron-density maps from MsbA-open-apo and -closed-apo.
Outward (Extracellular)-Facing Conformation of MsbA with Bound
Nucleotide. The structure of MsbA-AMPPNP has six transmem-
brane (TM) helices per monomer forming a dimer composed of 12
TM helices that is similar to Sav1866, with a rmsd of ?2.2 Å
between the C? positions of monomers (SI Fig. 8). The TM helices
extend into the cytoplasm and interact with the NBDs, providing
the link between the site of ATP hydrolysis and substrate transport
pathway. Intracellular helix 1 (IH1) (residues 112–120) is a short
helix between TM2 and TM3 that inserts down into a groove above
the P-loop (378–386) of the cis-NBD (SI Fig. 9) and makes contact
with the A-loop (351–358) and the nucleotide (cis, same monomer;
trans, opposite monomer). IH2 (residues 212–220) is a short helix
between TM4 and TM5 and is situated in a groove between the ?-
and ?-subdomains of the trans-NBD (SI Fig. 9). The helical TM6
providing the physical linkage between the TMD and NBD. Short
extracellular loops (EL1; residues 54–59, EL2; residues 162–166,
and EL3; residues 276–282) provide connections between TM1/
TM2, TM3/TM4, and TM5/TM6, respectively.
To further confirm the presence of nucleotide for this confor-
and vanadate to produce a transition state conformation with
bound ADP?Vi (48). The structure (MsbA-ADP?Vi) was solved by
molecular replacement (MOLREP) by using MsbA-AMPPNP.
4). Importantly, these crystals were obtained in the absence of
substrate, and the data are different from that previously reported
(B) Open apo. (C) Closed apo. One monomer in each model is colored with a
TM helices (TM1–TM6), extracellular loops (EL1–EL3), and intracellular helices
(IH1–IH2) are labeled accordingly. AMPPNP molecules are displayed as blue
sticks in the nucleotide-bound structure. In all structures, TM4/TM5/IH2 (yel-
low and orange) associates with the opposite monomer in a conserved
Stereoviews of three conformations of MsbA. (A) Nucleotide bound.
www.pnas.org?cgi?doi?10.1073?pnas.0709388104 Ward et al.
for MsbA in complex with ADP?Vi and LPS (39). The structures of
significant differences coming from either the bound mercury used
(Fig. 2C). A difference Fourier calculated by using model phases
(without nucleotide) between the ADP?Viand AMPPNP data sets
revealed that the top four positive peaks correspond to the ?
phosphate position (one per monomer), whereas the top eight
negative peaks correspond to the mercury positions (two per
monomer; SI Fig. 10A). A difference Fourier between the ADP?Vi
data and model with ADP alone also reveals the same vanadate
positions (one per monomer) as the top four peaks (SI Fig. 10B).
The position of the vanadate relative to the structural/catalytic
motifs of the ABC domain is consistent with a study of the maltose
ABC transporter (49).
MsbA from E. coli (MsbA-open-apo) in the absence of nucleotide
was built by using MsbA-AMPPNP as a starting point. A model of
a monomer was manually positioned into our experimental elec-
anomalous pairs (Fig. 3A). A single-copy model was refined by a
simulated annealing procedure (CNS1.2) by using two fourfold
NCS restraints (47) and refined to a Rcrystof 28% and Rfree31%
(Fig. 1B and SI Table 1) at 5.3-Å resolution and was validated by
generating a composite omit map (SI Movie 5). In this reanalysis,
we have identified density for the entire NBD, which is consistent
with x-ray structures of other isolated NBDs (reviewed in ref. 40).
Although the NBDs are ?50 Å apart, the ATP-binding half sites
face each other (Fig. 1B). This model may represent a functional
of the closed dimer due to the presence of detergent molecules
during crystallization cannot be excluded.
The most striking feature of this model is that TM4/TM5/IH2
crosses over and associates with the opposing monomer (Fig. 1B).
by the open-apo structure and provides the only means of interac-
tion between the two monomers. The association does not appear
to be a true domain swap, because it is difficult to model how
TM4/TM5/IH2 would interact with its own polypeptide in the same
manner as it interacts with its conjugate monomer. The cross-over
interaction buries 2,300 Å2surface area per monomer and likely
holds the dimer together during the open inverted V (inward-
facing) conformation. This type of interlocking mechanism, known
as an intertwined interface, has been observed in cytokines and
DNA-binding proteins and is thought to reinforce stability and
symmetry (50). Interestingly, the positions of the secondary struc-
ture elements TM4/TM5/IH2 are not equivalent in all copies of the
monomer in the asymmetric unit, suggesting that they move as a
rigid unit (SI Fig. 11).
Closed Inward (Cytoplasmic)-Facing Conformation of MsbA. The
structure of MsbA from V. cholerae (MsbA-closed-apo) was rebuilt
by using the experimental electron-density map calculated by using
the correct sign of the anomalous pairs (Fig. 3B). A model of
MsbA-closed-apo was built by using the model of MsbA-open-apo
several bulky side chains and confirm the registration of amino acids in the
mesh, contour 1?) showing the positions of Phe-116, His-214, Phe-349, Tyr-
351, Arg-360, Phe-392, Tyr-393, and Arg-416 in the NBD as yellow sticks. (D)
Stereoview of averaging omit density (white mesh, contour 1?) for AMPPNP
(green, blue, red, orange sticks) and the MsbA-AMPPNP model (white sticks).
Tyr-351 is shown as magenta sticks.
MsbA-AMPPNP dimer (C? model) contoured at 1? (blue ribbon) with solvent-
flattened and NCS-averaged density map (gray mesh) derived from MAD
experimental phases. The heavy-atom mercury locations (red surface) and
monomer. The unaccounted density is from symmetry-related molecules. (B)
Stereoview of side-chain electron density for MsbA-AMPPNP at 3.7-Å resolu-
tion derived from an averaging omit map phased from a model excluding
nucleotide and truncated to polyalanine/glycine (white mesh, contour 1?).
The stacking of Phe-105 and -230 residues on adjacent TM helices and density
for Arg-102, Met-109, Met-210, and His-214 are shown as yellow sticks. The
initial phases for the averaging omit map were derived from a full model
excluding nucleotide and truncated to polyalanine/glycine. The initial unav-
eraged map (blue mesh, contour 1?) using polyalanine/glycine model phases
is shown for reference. The recovered densities for the side chains are a
Electron-density maps for MsbA-AMPPNP. (A) Stereoview of a single
Ward et al. PNAS ?
November 27, 2007 ?
vol. 104 ?
no. 48 ?
previous retracted model, fragments of the NBD were overfitted
into an incorrect map calculated in the wrong hand. A single-copy
model was refined to a Rcrystof 35% and Rfree36% (Fig. 1C and SI
Table 1) at 5.5-Å resolution. The final model was also validated by
generating a composite omit map (SI Movie 6). Despite the lower
resolution of the structure, the molecular envelope of the protein
and helices is well resolved (Fig. 3B). Further, the topology of the
TMD was verified by mercury positions Cys-88 and -315, and the
placement of the NBD in the electron density was confirmed by a
mercury bound to Cys-401 (Fig. 3B). This structure, like MsbA-
open-apo, is without nucleotide and represents another possible
Structural and biochemical studies of isolated ABC domains of
Rad50, MJ0796, MJ1267, and MalK have confirmed a mechanism
by which the NBDs of ABC transporters coalesce in an ATP-
dependent manner and undertake nucleotide hydrolysis (5, 6, 15,
17, 51–53). From these experiments alone, it is unclear whether the
NBDs remain spatially together during the transport cycle, or to
what extent these half-sites could come apart in the context of the
full-length protein. Although these and other biochemical experi-
ments have proven to be invaluable tools for probing function, the
range of motions that can be inferred from these types of studies is
to probe the steps of the transport cycle are necessarily restricted to
the nucleotide-bound state. In this work, we provide structural
evidence showing that MsbA can undergo a large range of motions
between the apo- and nucleotide-bound states, whereas the cata-
lytic ABC half-sites remain facing each other.
homologous importers, have two different orientations (inward-
and outward-facing), but the conformational changes between
them are small. These importers transport specific molecules into
their extracellular face. In contrast, MsbA and Sav1866 (22, 54) are
homologous exporters that do not interact with cognate proteins.
Unlike importers, MsbA and Sav1866 likely recruit substrate dif-
fusing along the inner leaflet of the membrane through lateral
interactions and/or directly from the cytoplasm. In addition, MsbA
functions natively to transport LPS, which may require significant
the membrane. Nucleotide-bound structures of Sav1866 and MsbA
provide a snapshot of the outward-facing conformation of an ABC
exporter. A comparison of the outward-facing conformation with
the inward-facing conformations highlights the flexibility and large
range of motions possible for this class of proteins.
The conformational changes between the different structures of
MsbA are most apparent when monomers from each state are
aligned by using transmembrane (TM) helices 1, 2, 3, and 6 (rmsd
formed by EL2/EL3 in a rigid-body motion with nearly conserved
NBD alignment (SI Movie 1). This hinge creates an inward-facing
apo conformations of MsbA. The NBDs in the closed apo state,
although closer, have not formed an ATP sandwich, and the
P-loops of opposing monomers are positioned next to one another
(Fig. 5C). This P-loop arrangement is consistent with cross-linking
studies of MDR1 (55, 56) showing that cysteines in the P-loops are
of nucleotide (SI Fig. 12). This could happen only if the dimer
arranged similarly to the closed-apo inward-facing conformation of
MsbA. In addition, the reaction occurred only at 37°C for MDR1,
the residues close enough to react (56).
The conformational transition from MsbA-closed-apo to MsbA-
AMPPNP results in a tight canonical NBD interface with nucleo-
(Fig. 5C). This transition can be described as two unique motions
but likely occurs as a concerted movement. The first motion is a
?10° pivot about the hinge, EL2/EL3 (Fig. 4) described above,
bringing the NBDs closer but not into alignment. Next, the NBDs
single MsbA-open-apo dimer (C? model) contoured at 1? (blue ribbon) with
solvent-flattened and NCS-averaged density map (gray mesh) derived from
MAD/SIRAS combined experimental phases. (B) Stereoview of a single MsbA-
closed-apo dimer (C? model) contoured at 1? (blue ribbon) with solvent-
flattened and NCS-averaged density map (gray mesh) derived from MAD
experimental phases. The heavy-atom Hg positions (red surface) and corre-
sponding C? positions (yellow balls) of Cys-88, -315, and -401 are superim-
one monomer. The unaccounted density is from symmetry-related molecules.
Electron-density maps for MsbA apo structures. (A) Stereoview of a
mational changes within the MsbA monomer demonstrating the mobility of
on TM helices 1, 2, 3, and 6. Two NCS-related monomers of MsbA-open-apo
that differ in the relative positioning of TM4/TM5/IH2 are shown in blue and
cyan (see also SI Fig. 11). The monomers from MsbA-closed-apo and MsbA-
AMPPNP are shown in green and yellow, respectively. IH1 moves very little,
whereas TM4/TM5/IH2 moves significantly.
Comparison of the MsbA monomers. (A) Stereoview of the confor-
www.pnas.org?cgi?doi?10.1073?pnas.0709388104Ward et al.
dimerize and form a canonical ABC interface. This second motion
tilts the TM4/TM5 pair ?20° out of plane (relative to the first
motion) about a pivot point centered along its length (SI Movie 1).
Because TM3/TM6 are connected to TM4/TM5 via EL2 and EL3,
the formation of the NBD sandwich pulls TM3/TM6 away from
TM1/TM2 (Fig. 5 A and B). The newly formed outward opening is
the inward opening formed between TM3/TM6 and TM4/TM5.
Thus, changes in both the orientation and spacing of the NBDs
dramatically rearrange the packing of transmembrane helices and
effectively switch access to the internal chamber from the inner to
the outer leaflet of the bilayer. These mechanical motions are
illustrated in SI Movie 1 and paper cutout model (SI Fig. 13).
The conformational changes in MsbA are supported by struc-
tural studies of the ABC transporter MalK, which nicely illustrate
how the NBDs associate in a nucleotide-dependent fashion (17). In
MalK, the NBDs come apart in the absence of nucleotide, but the
catalytic half-sites remain facing each other, because an additional
C-terminal domain prevents their complete dissociation. By anal-
ogy, the NBDs of MsbA can also separate with the ABC-binding
half-sites opposing each other but are held together by the TMs
(Fig. 1 B and C). The open-apo conformation of other ABC
transporters has also been observed in structural studies of BmrA
disassociation of NBDs is also consistent with the x-ray structures
of several isolated ABC domains that do not form the ATP
sandwich dimer in the absence of nucleotide (reviewed in ref. 40).
Although the NBDs are separated in the open-apo conforma-
tion, the large inward-facing opening may serve an important
biological role. MsbA and homologous transporters like MDR1
pump substrates out of the cell by first recruiting them from the
inner leaflet of the bilayer and/or the cytosol. In the case of MsbA,
LPS is first synthesized on the inner leaflet and then transported
across the inner membrane (28). From the nucleotide-bound
structures of MsbA or Sav1866 alone, it is difficult to reconcile how
membrane leaflet side, much less from the cytoplasm. These
conformations have no aqueous opening facing the cytoplasm, and
open- and closed-apo structures, however, are inward-facing, and
these conformations can accommodate substrate from either the
to the inward-facing conformation (modeled in SI Fig. 16) may
promote the closure of the transporter’s TMDs, which would, in
turn, reposition the NBDs, allowing the formation of the ATP
sandwich in the presence of nucleotide.
The most convincing study supporting the open V shape con-
formation of MsbA is from electron paramagnetic spin resonance
(EPR). EPR studies (42) predicted a large change in the aqueous
is in close agreement with the structural transition from open-apo
to nucleotide-bound. Cross-linking studies of lipid membrane-
embedded MsbA (59) show that a cysteine mutation at residue 56
forms a spontaneous disulphide with the opposing monomer. The
open-apo structure is the only conformation that brings these two
residues close to the required distance for this reaction to occur.
Although still too far apart for spontaneous disulphide bond
formation, the C? positions of residue 56 (?15 Å) in this open-apo
structure are much closer than that observed in the nucleotide-
bound conformation (?35 Å) (SI Fig. 17). Flexibility in EL1 or
further opening could bring the residues even closer.
of sampling a large conformational space. The dimer interface in
the inward-facing conformations of MsbA is mediated exclusively
by the intertwined interface (TM4/TM5/IH2) with the opposing
monomer. In the absence of nucleotide, the large buried surface
area contributed by this interface stabilizes the dimer, allowing
flexibility in the transporter while maintaining the relative orien-
tation of the individual domains. The relative positioning of TM4/
TM5/IH2 to the opposing monomer is structurally preserved in
each conformation of MsbA (Fig. 1). In the apo state, the trans-
porter may sample different conformations, but the NBDs cannot
In the nucleotide-bound state, the NBDs come together to form a
canonical ATP dimer sandwich, significantly increasing the molec-
ular interface within the protein. This dimerization of the NBDs is
coupled to a packing reorganization of the TM helices relative to
the nucleotide-free state. The resultant twisting motion pulls TM3/
TM6 away from TM1/TM2 and causes a change from an inward-
to an outward-facing conformation (SI Movie 1). In this ‘‘alternat-
ing access’’ (60) model of MsbA, the inward and outward openings
are mediated by two different sets of TM helices (TM3/TM6 and
TM4/TM5 vs. TM3/TM6 and TM1/TM2, respectively). The differ-
changes within the MsbA dimer alter the accessibility to the internal chamber
from inward to outward facing. For clarity, only TM helices (labeled 1–6) of
one monomer (cyan) are shown inside a surface rendering of the dimer. The
open and closed apo conformations form an inward-facing V between TM4/
AMPPNP) forms an outward-facing V between TM3/TM6 and TM1/TM2, just
moves, causing TM3/TM6 to split away from TM1/TM2, which results in an
outward-facing conformation. Both inward- and outward-facing conforma-
tions are mediated by intramolecular interactions within a single monomer,
but by different sets of helices. (B) Simplified cartoon model illustrating the
points above. The relative position of each TM helix is labeled with a number
(one monomer in white and the other in gray). The arrows illustrate the
motions required to go to the next state. (C) Top-down view of NBDs (one
monomer shown in white and the other in gray). IH1 (green) and IH2 (yellow)
facing each other; the P-loops (red) are roughly aligned (dashed lines) with
one another across the dimer interface. Upon nucleotide binding (AMPPNP -
magenta), the canonical ATP sandwich is formed, aligning the nucleotide
tracks with the trans-monomer. The motion of the NBDs from closed-apo- to
nucleotide-bound transmits a structural change (described above) to the TMs
via IH1 and IH2, resulting in an outward-facing conformation.
Summary of conformational changes in MsbA. (A) Conformational
Ward et al.PNAS ?
November 27, 2007 ?
vol. 104 ?
no. 48 ?
across the membrane, especially if the outward-facing conforma-
tion has lower affinity for substrate (61–63).
about the extent and the biological role of flexibility in ABC
transporters in the context of the ATP switch model (64, 65). The
structures, however, are only static representations and must be
further investigated by other means. Cross-linking studies, FRET,
and EPR are valuable tools for making distance measurements
between residues and calculating the solvent accessibility of resi-
dues. Together with structural studies, these techniques will help
tackle the challenges of studying the complex range of motion of a
Materials and Methods
the pET19b expression vector and overexpressed by using E. coli
host BL21 (DE3) grown by fermentation. Solubilized MsbA was
anion-exchange, and gel-filtration chromatography. ATPase activ-
ity was measured by using a linked enzyme ATP-regenerating
system. Crystals were grown at 4°C by using the sitting-drop
vapor-diffusion method. CNS1.2 was used for crystallographic
refinement. For more details, see SI Text.
We thank Drs. D. C. Rees, R. Milligan, P. Wright, H. van Veen, and I.
Science Foundation Minority Postdoctoral Fellowship. A.W. is sup-
ported by the Norton B. Gilula Fellowship. This work was supported by
the National Institutes of Health (Grants GM61905 and Roadmap
GM073197), the U.S. Army (Grant W81XWH-05-1-0316), the National
Aeronautic and Space Administration (Grant NAG8-18334), the Beck-
man Foundation, the Fannie E. Rippel Foundation, Skaggs Chemical
Biology, and the Baxter Foundation.
1. Higgins CF (1992) Annu Rev Cell Biol 8:67–113.
2. Paulsen IT, Sliwinski MK, Saier MH (1998) J Mol Biol 277:573–592.
3. Schneider E, Hunke S (1998) FEMS Microbiol Rev 22:1–20.
4. Holland IB, Blight MA (1999) J Mol Biol 293:381–399.
5. Hopfner KP, Karcher A, Shin DS, Craig L, Arthur LM, Carney JP, Tainer J
(2000) Cell 101:789–800.
6. Davidson AL (2002) J Bacteriol 184:1225–1233.
7. Gottesman MM, Pastan I, Ambudkar SV (1996) Curr Opin Genet Dev
8. Gottesman MM, Ambudkar SV (2001) J Bionenerg Biomembr 33:453–458.
9. Dean M (2005) Methods Enzymol 400:409–429.
10. Dean M, Annilo T (2005) Annu Rev Genomics Hum Genet 6:123–142.
11. Biemans-Oldehinkel E, Doeven MK, Poolman B (2006) FEBS Lett 580:1023–
12. Hung LW, Wang IX, Nikaido K, Liu PQ, Ames GF, Kim SH (1998) Nature
13. Karpowich N, Martsinkevich O, Millen L, Yuan YR, Dai PL, MacVey K,
Thomas PJ, Hunt JF (2001) Structure (London) 9:571–586.
14. Yuan YR, Blecker S, Martsinkevich O, Millen L, Thomas PJ, Hunt JF (2001)
J Biol Chem 276:32313–32321.
15. Smith PC, Karpowich N, Millen L, Moody JE, Rosen J, Thomas PJ, Hunt JF
(2002) Mol Cell 10:139–149.
16. Locher KP, Lee AT, Rees DC (2002) Science 296:1091–1098.
17. Chen J, Lu G, Lin J, Davidson AL, Quiocho FA (2003) Mol Cell 12:651–661.
18. Schmitt L, Benabdelhak H, Blight MA, Holland IB, Stubbs MT (2003) J Mol
19. Karcher A, Buttner K, Martens B, Jansen RP, Hopfner KP (2005) Structure
20. Lu G, Westbrooks JM, Davidson AL, Chen J (2005) Proc Natl Acad Sci USA
21. Zaitseva J, Jenewein S, Jumpertz T, Holland IB, Schmitt L (2005) EMBO J
22. Dawson RJ, Locher KP (2006) Nature 443:180–185.
23. Kitaoka S, Wada K, Hasegawa Y, Minami Y, Fukuyama K, Takahashi Y (2006)
FEBS Lett 580:137–143.
24. Pinkett HW, Lee AT, Lum P, Locher KP, Rees DC (2007)Science315:373–377.
25. Karow M, Georgopoulos C (1993) Mol Microbiol 7:69–79.
26. Polissi A, Georgopoulos C (1996) Mol Microbiol 20:1221–1233.
27. Zhou Z, White KA, Polissi A, Georgopoulos C, Raetz CR (1998) J Biol Chem
28. Doerrler WT, Reedy MC, Raetz CR (2001) J Biol Chem 276:11461–11464.
29. Doerrler WT, Raetz CR (2002) J Biol Chem 277:36697–36705.
30. Doerrler WT, Gibbons HS, Raetz CR (2004) J Biol Chem 279:45102–45109.
31. Wang X, Karbarz MJ, McGrath SC, Cotter RJ, Raetz CR (2004) J Biol Chem
32. Reuter G, Janvilisri T, Venter H, Shahi S, Balakrishnan L, van Veen HW
(2003) J Biol Chem 278:35193–35198.
33. Woebking B, Reuter G, Shilling RA, Velamakanni S, Shahi S, Venter H,
Balakrishnan L, van Veen HW (2005) J Bacteriol 187:6363–6369.
34. Shilling RA, Venter H, Velamakanni S, Bapna A, Woebking B, Shahi S, van
Veen HW (2006) Trends Pharmacol Sci 4:195–203.
35. Van Veen HW, Callaghan R, Soceneatu L, Sardini A, Konings WN, Higgins
CF (1998) Nature 391:291–295.
36. Chang G, Roth CB, Reyes CL, Pornillos O, Chen YJ, Chen AP (2001) Science
293:1793–1800, and retraction (2006) 314:1875.
37. Chang G, Roth CB (2001) Science 293:1793–1800.
38. Chang G (2003) J Mol Biol 330:419–430.
39. Reyes CL, Chang G (2005) Science 308:1028–1031.
40. Oswald C, Holland IB, Schmitt L (2006) Naunyn Schmiedebergs Arch Phar-
41. Higgins CF, Linton KJ (2001) Science 293:1782–1784.
42. Dong J, Yang G, McHaourab HS (2005) Science 308:1023–1028.
43. Urbatsch IL, Gimi K, Wilke-Mounts S, Senior AE (2000) J Biol Chem
44. Steinfels E, Orelle C, Fantino JR, Dalmas O, Rigaud JL, Denizot F, Di Pietro
A, Jault JM (2004) Biochemistry 43:7491–7502.
45. Furey W, Swaminathan S (1997) in Methods in Enzymology: Macromolecular
Crystallography (Academic, Orlando, FL), Vol 277, Part B, Chap 31.
46. Sack JS (1988) J Mol Graphics 6:224–225.
47. Bru ¨nger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve
RW, Jiang JS, Kuszewski J, Nilges M, Pannu NS, et al. (1998) Acta Crystallogr
48. Smith CA, Rayment I (1996) Biochemistry 35:5404–5417.
49. Fetsch EE, Davidson AL (2002) Proc Natl Acad Sci USA 99:9609–9610.
50. Larsen TA, Olson AJ, Goodsell DS (1998) Structure (London) 6:421–427.
51. Davidson AL, Laghaeian SS, Mannering DE (1996) J Biol Chem 271:4858–
52. Davidson AL, Sharma S (1997) J Bacteriol 179:5458–5464.
53. Moody JE, Millen L, Binns D, Hunt JF, Thomas PJ (2002) J Biol Chem
54. Dawson RJ, Locher KP (2007) FEBS Lett 581:935–938.
55. Loo TW, Clarke DM (2000) J Biol Chem 275:19435–19438.
56. Urbatsch IL, Gimi K, Wilke-Mounts S, Lerner-Marmarosh N, Rousseau ME,
Gros P, Senior AE (2001) J Biol Chem 276:26980–26987.
57. Chami M, Steinfels E, Orelle C, Jault JM, Di Pietro A, Rigaud JL, Marco S
(2002) J Mol Biol 315:1075–1085.
58. Hofacker M, Gompf S, Zutz A, Presenti C, Haase W, van der Does C, Model
K, Tampe R (2007) J Biol Chem 282:3951–3961.
59. Buchaklian AH, Funk AL, Klug CS (2004) Biochemistry 43:8600–8606.
60. Jardetzky O (1966) Nature 211:969–970.
61. Ramachandra M, Ambudkar SV, Chen D, Hrycyna CA, Dey S, Gottesman
MM, Pastan I (1998) Biochemistry 37:5010–5019.
62. Rosenberg MF, Velarde G, Ford RC, Martin C, Berridge G, Kerr ID,
Callaghan R, Schmidlin A, Wooding C, Linton KJ, Higgins CF (2001) EMBO
63. Loo TW, Clarke DM (2002) Proc Natl Acad Sci USA 99:3511–3516.
64. Higgins CF, Linton KJ (2004) Nat Struct Mol Biol 10:918–926, review.
65. Higgins CF, Linton KJ (2007) Pflu ¨gers Arch 453:555–567.
www.pnas.org?cgi?doi?10.1073?pnas.0709388104 Ward et al.